Watermark techniques have been heretofore used for determining whether a bill and the like are authentic or counterfeit. In addition, these years, as computer technologies are being developed, there have been an increased number of cases where music, an image, an animation are supplied as digital contents on a basis of copyrights thereof. For this reason, an unauthorized copy of the contents needs to be prevented by using the aforementioned “watermark” techniques in order to determine whether the contents have been copied in an unauthorized manner. Embedding of a “watermark” in contents is performed through embedding a “watermark signal” (hereinafter referred to as a “wartermark signal”) in original contents in a digital manner as a general practice.
Various digital watermark embedding methods have been heretofore proposed. For example, in “a system and method for embedding data in a frequency domain” proposed by the same applicant, it has been considered that, when a digital watermark is intended to be embedded in contents including audio signals such as music, a psychoacoustic model is calculated in the frequency domain, thereby embedding the digital watermark. In this method, a DFT (discrete Fourier Transform) frame needs to be detected exactly when a frequency of an audio signal is detected. This increases time needed for the calculation. For this reason, this method has a disadvantage that the method is not suitable for a purpose of embedding, without causing time delay, wartermark signals in audio signals being supplied in a real time manner.
With this taken into consideration, in a patent application filed by the same applicant, entitled “a system and method for embedding a watermark without requiring frame synchronization,” a technique of embedding wartermark signals without requiring a frame to be synchronized with audio signals has been considered. The embedding technique without requiring the frame synchronization has the following advantages. First, robustness to expansion, contraction and locational shift of original signals is large. Second, the time delay is not so large. Third, the digital watermark can be detected and judged with good performance. However, with regard to the aforementioned embedding technique without requiring the frame synchronization, phases respectively of the wartermark signals and the original signals need to be synchronized. For this reason, the embedding technique without requiring the frame synchronization is not suitable for a purpose of embedding wartermark signals in contents being supplied in a real time manner, from which only the wartermark signals are delayed. Accordingly, the embedding technique without requiring the frame synchronization has had a trouble in robustness. That is, when delay is caused only in wartermark signals, capability of detecting the wartermark signals is deteriorated to a large extent.
In addition, according to Boney, et al, “Digital Watermarks for Audio Signals,” (IEEE International Conference on Multimedia Computing and System, Jun. 17-23, 1996, Hiroshima, Japan, pp. 473-480), the aforementioned trouble of deteriorated robustness due to delayed wartermark signals is corrected by providing a filter for simulating the psychoacoustic model in advance and by filtering a pseudo-random sequence in the time domain. However, a filtering coefficient has to be determined for each frame. For this reason, in common with the above mentioned method, this method is not suitable for embedding wartermark signals in audio signals being supplied in a real time manner.
Furthermore, in “Robust Audio Watermarking Using Perceptual Masking” (Signal Processing, Vol. 66, 1998, pp. 337-355), Swanson and his group have proposed a method using both a psychological audio-visual sensation model for calculating frequency masking and a psychoacoustic model for calculating temporal masking. With regard to the temporal masking, wartermark signals are embedded by prediction of an envelope calculation of audio signals and an amount of masking. However, the prediction is a prediction of an amount of masking by use of an output of the temporal masking. The method is not for embedding wartermark signals in a real time manner by directly using the original contents.
Apart from this, a technique has been proposed of embedding a digital watermark in image data such as video signals. For example, in “Robust 3D DFT Video Watermarking,” Proc. SPIE, Vol. 3657, pp. 113-124, 1999, Deguillaum and his group have proposed a method for embedding wartermark signals by adapting a DFT with a video sequence used as three-dimensional information constituted of a vertical, horizontal and temporal axes. Even in this method, delay is caused by a time width for which the DFT is performed. For this reason, this method is not suitable for embedding a digital watermark in a real time manner.
Moreover, Japanese Patent Laid-open Official Gazette No. Hei. 11-55638 has disclosed a method which defines a partial area in an image as an area to which information is added, and which embeds the information in the image by enlarging or reducing this area. This embedding method does not add a wartermark signal to the image, but processes a part of the image itself. For this reason, a difference between a pre-embedded signal and a post-embedded signal is so large that a problem with quality is brought about. In addition, in “Watermarking of Uncompressed and Compressed Video,” Signal Processing, Vol. 66, No. 3, pp. 283-301, 1998, Hartung and his group have disclosed a method which regards a video as a continuation of still images, and which adds a message which has been modulated by a pseudo-random sequence to each frame. Additionally, Hartung and his group have proposed a method in which a compressed video sequence is not decoded, and in which a DFT count is replaced depending upon a message whenever deemed necessary. However, even the method proposed by Hartung and his group does not perform a control in a predictive manner. In this point, the method is not satisfactory in embedding wartermark signals in contents in a real-time manner.